A test with Icon Solutions hit 6,000 transactions per second with zero data loss.
MongoDB and fintech company Icon Solutions published joint benchmark results on 8 September 2026. The testing demonstrated that the Icon Payments Framework running on MongoDB Atlas processed 6,000 end-to-end payment transactions per second with a mean latency of 0.51 seconds.
The 6,000 transactions per second workload required around 430,000 database operations per second across two MongoDB Atlas clusters. About 236,000 operations ran on the event journal and 196,000 on the operational data store. During peak load, hot-shard central processing unit utilisation on MongoDB Atlas remained at around 50%, while the WiredTiger write queue stayed at zero.
The tests also evaluated platform resilience under live transaction traffic. A MongoDB Atlas primary node failover finished in about 5 seconds on average with no payment loss or application restarts. Upgrading from MongoDB Atlas version 8.0 to 8.3 cut persistence latency by approximately 50%, falling from around 40 milliseconds to roughly 20 milliseconds at 4,000 transactions per second.
Boris Bialek, Vice President of Industries and Global Field CTO at MongoDB, stated that instant payments require an event-driven data foundation rather than a static system of record, noting the platform handled the load with headroom to spare.
Newsletter
Markets in your inbox, weekly
Latin America-focused analysis, investment themes and the week in finance.
Keep reading